Implicit conversion from float -102.94736842105263 to int loses precision